To reduce margins, researchers are developing methods for in vivo proton range verification. Two approaches – PET imaging of positron emitters generated in the patient and MR imaging of radiation-induced tissue changes – have been demonstrated clinically. Now, researchers from the University of Pennsylvania and IBA have reported the first clinical use of prompt gamma imaging (PGI) for range verification in pencil-beam scanned (PBS) proton therapy (Int. J. Radiat. Oncol. Biol. Phys. doi: 10.1016/j.ijrobp.2017.04.027).

PGI works by detecting the prompt gamma rays created when the incident proton beam interacts with atomic nuclei in the body. The researchers employed a prototype prompt gamma camera, developed by IBA, to image proton tracks by measuring 1D profiles of prompt gammas transmitted through the patient. As prompt gamma rays are emitted less than 1 ns after excitation, each gamma photon in the signal can be associated with a particular pencil-beam spot.

The gamma camera is based on two rows of 20 LYSO scintillator crystal slabs, with each slab coupled to an array of silicon photomultipliers. Each 31.5 x 100 x 4 mm slab and its mirror slab constitutes one of 20 bins in the detection profile. The prompt gammas enter the camera through a 6 mm knife-edge slit aperture in a tungsten collimator and produce a reversed projection of their depth profile on the crystals.

The geometry, optimized with researchers from Université Libre de Bruxelles (see: Can prompt gammas monitor in real time?), favours high detection efficiency and benefits from electronics (designed by partners at Politecnico di Milano and XGLab) that can handle the huge corresponding count rate with negligible dead time.

IBA has built two prototype systems, with the other in use since 2015 at OncoRay in Dresden, where the first PGI patient study in double-scattering mode was performed (see: Prompt gamma imaging goes clinical). The University of Pennsylvania team is the first to use the PGI system clinically in a pencil-beam scanning proton treatment. "We have performed PGI in two patients so far, and we plan to do a larger scale clinical trial," explained project supervisor Kevin Teo.

Millimetre precision

Researcher Yunhe Xie and colleagues used the camera to record the prompt gamma signal emitted during pencil-beam scanned proton therapy in a brain cancer patient. They performed PGI for six fractions over three weeks of treatment, delivered via three fields (right lateral, right-superior-oblique and superior-inferior), each comprising 17 or 18 energy layers. "The system performs time-resolved acquisition, so it can record the position of each spot in each layer separately," Teo explained.

To determine shifts in Bragg peak depth, the researchers compared the measured prompt gamma depth profiles of individual pencil-beam spots with expected profiles simulated from the treatment plan. The mean range shifts aggregated over all spots in nine energy layers were –0.8±1.3 mm for the lateral field, 1.7±0.7 mm for the oblique field and –0.4±0.9 mm for the superior-inferior field. Range shifts were, in this case, verified to be smaller than the 5 mm distal margin applied clinically.

The accuracy of the computed shift is highly dependent upon the statistical noise in the profiles, with better precision achieved in spots with more protons. "In practice, because the spots are delivered on a fine scanning grid and are substantially overlapping to build a homogeneous dose distribution, many of them don't deliver much dose and produce a weak signal," noted Teo. Aggregating the signal from neighbouring, overlapping spots can increase the signal-to-noise ratio. While this will also degrade the lateral spatial resolution, the researchers note that spot aggregation with 4–7 mm sigma has a minimal impact, reducing the spatial resolution from 5 to 8.6 mm.

For all spots, the median precision in shift estimation after spot aggregation with a kernel of 7 mm sigma was 2.1 mm, and the median inter-fraction variation was 1.8 mm. Without aggregation, the precision was better than 2 mm for only 1% of the 5801 analysed spots. Aggregating spots with 4 and 7 mm sigma resulted in 13% and 46% of spots with precision better than 2 mm, respectively. With a kernel of 7 mm sigma, 52 spots were retrieved with a precision better than 1 mm.

The accuracy of the system for range verification is limited by the positioning accuracy of the system. The researchers note that positioning accuracy could be increased further. The present, standalone, trolley positioning system was built to enable this study with no impact to the treatment room. Positioning accuracy, as well as distance to the beam axis, could be further improved through tighter integration of the camera system within the therapy system.

Once further integrated, PGI holds promise for many potential clinical applications. PGI could be used, for example, as a quality assurance (QA) tool to reduce range uncertainty margins, for patient-specific correction of shifts in proton stopping power calibration, or to help enable future online adaptive treatment workflows. PGI spot shift data would also permit online QA of the adapted plan through reconstruction of the delivered dose.

"Right now there are no in vivo tools available to us, this is just a starting point," Teo told medicalphysicsweb. "I can see applications where we would consider using PGI routinely, stereotactic treatments, for example, where you deliver more dose per fraction. If the target is close to sensitive structures this would be very useful."

Related articles

• Prompt gamma imaging goes clinical
• Ranging scheme may improve proton therapy
• Resolved gamma lines verify proton range
• Can prompt gammas monitor in real time?